The transparent optical network (TON) is an attractive network paradigm offering high data rates without expensive O-E-O conversion, and will be more available to public users as fiber-to-the-home (FTTH) getting increasingly popular. However, transparency will also introduce attack threats to TON, e.g., malicious users can gain more chances to access to the network, and then inject a beam of light at a high power being 20 dB or even higher than a normal one, which will result in crosstalk attack on normal signals.1, 2, 3, 4, 5 Especially in the optical switch architectures, such as optical cross connect (OXC), a high-powered attack signal will leak significant power to normal channels working at the same wavelength, resulting in intrachannel crosstalk attack.
A model to describe the propagation of intrachannel crosstalk attack in a TON is proposed as shown in Fig. 1.1, 3 In this figure, a high-powered signal (Attacker) will leak power to the legitimate signal (User 1) through intrachannel crosstalk, and such significant leakage will enable User 1 the attack capability. The attack capability in User 1 will in turn affect User 2 in the next switch, therefore more stages of switches will be affected.1, 3, 5 The high-powered signal will also rob the gain of adjacent channels in an optical amplifier and become stronger,4, 6 which will make the attack propagation of intrachannel crosstalk more serious.
In this letter, we present three scenarios of intrachannel crosstalk attack, attack propagation within the first OXC, the secondary attacker traversing successive OXCs, and the original attacker traversing successive OXCs. Via VPItransmakerTM, analysis on bit-error-rate (BER) and eye diagram penalties imposed by attack signals with different switch crosstalk intensities and detection methods are presented. The penalties accompanied with gain competition attack are also given. The simulation proved that the original attacker will cause the propagation effect of intrachannel crosstalk attack within three successive OXCs but with a limited two stages of switches in each OXC. The results also proved that the polluted signals (i.e., the polluted secondary attacker) do not have enough attack capabilities to propagate intrachannel crosstalk attack to the next OXC. The simulation also indicates that if gain competition attack exists, the BER will be somewhat higher.
Simulation Analysis and Setup
Figure 2 depicts the simulated TON system, in which four wavelengths are multiplexed in a fiber and all four laser transmitters (TxExtModLaser) transmit at a power of 1 mW. All the transmitted signals are nonreturn to zero non-return-to-zero (NRZ) formats at a rate of 10 Gbit/s and modulated on four wavelength channels λ0, λ1, λ2, and λ3 according to ITU 100 GHz grid from 193.00 to 193.30 THz at C-band. The grid spacing is wide enough to suppress four-wave mixing and cross-phase modulation (XPM)6 to focus on intrachannel crosstalk attack. However, the extremely high-powered signal will cause serious self-phase modulation (SPM), which causes the broadening of the signal spectrum and power degradation of the attack signal.7 The statistics of phase-difference between legitimate and crosstalk signal which dominates the BER performance8 is also difficult to determine due to serious SPM, thus multisimulations are carried out by adjusting phase shift in optical switch and we select the worst BER. All segment fibers are nonlinear dispersive fibers (NLS) with 0.2 dB/km attenuation and 2.6 × 10−20 m2/W nonlinear index. The dispersion for all NLS segments is set identical 2.0 × 10−3 ps/nm/km also without compensation so that we can concentrate on the crosstalk attack. We employ erbium doped fiber amplifier (EDFA) (AmpSysOpt) as an amplifier with 16 dB fix gain to avoid gain competition attack and 4 dB noise figure. Multiplexer and switch are with 2 dB insertion loss. Node spacings are set to 400 km including five loops of 80 km NLS segments.9
We assume that the attacker injects a high-powered signal modulated at the same 10 Gbit/s NRZ format as legitimate signals on wavelength λ1 at a power of 500 mW and the injection point is 15 km before OXC-1.2 In each OXC, three cascaded 2 × 2 optical switches (SwitchDos-Y-Two) are set for channel λ1. Let λ1, n represent the legitimate signal on wavelength λ1 passing the n’th stage of an optical switch. The high-powered attacker on wavelength λ1 will attack legitimate signals λ1, 1, λ1, 2 and λ1, 3 from OXC-1 to OXC-6, as shown in Fig. 2. The polluted signal in OXC-1 directly traverses OXC-7 and OXC-8. At the egress points of each switches (i.e., the point labeled as @), BERs and eye diagrams are detected using RxBERs and ViScopes.
Simulation and Discussion
Intrachannel Crosstalk Attack Propagation Within OXC
We assume the crosstalk to be Gaussian distribution to achieve the upper bound of system BER (Ref. 10) and the statistics of the received optical signal will follow one of a family of Chi-squared probability densities,11, 12 and then the Chi-squared method is set to estimate the BER at the receivers. As a comparison, the Gaussian method is also implemented, in which the statistics of the received optical signal are assumed to be Gaussian distribution. Switch crosstalk intensity is a parameter representing the amount of power leakage between two channels in an optical switch, which determines how much crosstalk noise will be added in the received optical signals. Figure 3 illustrates the BER of signals λ1, 1, λ1, 2, and λ1, 3 in OXC-1. As switch crosstalk intensity decreases from −20 to −35 dB in four grades, i.e., −20, −25, −30, and −35 dB,4 the BERs for λ1, 1 are affected between 0.1 and 0.5, and those for λ1, 2, are distinguishingly distributed. However, BERs for λ1, 3 are all kept lower. Eye diagrams for λ1, 2 with different switch crosstalk intensities under Chi-squared detection are also given in Fig. 4. The result reveals that the extent of intrachannel attack propagation within an OXC is different with switch crosstalk intensities but with limited two stages of switches. The result also indicates that as the crosstalk increases, there will be less difference between detection methods. Compared to the Gaussian method, the Chi-squared method will overestimate the system BER.
Original Attacker Traversing Successive OXCs
To investigate the intrachannel crosstalk attack spanning multiple OXCs, the original attack signal is set to traverse from OXC-1 to OXC-6. The switch crosstalk intensity is set to be identically −25 dB at all switches. Figure 5 shows the BERs of λ1, 1, λ1, 2, and λ1, 3 in OXC-2 to OXC-6, respectively. The BERs of λ1, 1 at OXC-2 and OXC-3 are 0.3 and 1.05 × 10−5, respectively. The BER of λ1, 1 at OXC-4, -5, and -6 are all quickly dropped. Those BERs of λ1, 2 and λ1, 3 at five OXCs are all less than 1.0 × 10−10. Eye diagrams of λ1, 1 in OXC-2 to OXC-5 are shown in Fig. 6. By setting EDFA to power model, a gain competition effect can be achieved. Figure 5 shows the BER penalties accompanied with gain competition attack, in which we can see that the robbed gain in EDFA will make the BER of λ1, 1 in OXC-2 to OXC-6 a little worse for the reason that −25 dB of the robbed power will be leaked to the legitimate channels. The results also indicate that only the first stage of switches will be affected worse if gain competition attack exists. The results indicate that the high-powered original intrachannel crosstalk can propagate its attack effect to successive three OXCs and the BER will be somewhat higher in case there is gain competition attack.
Secondary Attacker Traversing Successive OXCs
The intrachannel crosstalk propagation caused by the polluted signal (secondary attacker) is also simulated. As illustrated in Fig. 2, the secondary attacker from OXC-1 traverses OXC-7 and OXC-8 and attacks λ1, 1, λ1, 2, and λ1, 3 in the OXCs with BERs all being below 1.0 × 10−9 as shown in Fig. 5. If gain competition attack exists, as shown in Fig. 5, the BER of λ1, 1 in OXC-7 and OXC-8 are getting a little higher. The simulation indicates that the polluted secondary attacker does not have enough capabilities to propagate intrachannel attack to successive OXCs even if gain competition attack exists, for the reason that insertion loss in optical components and nonlinear effect in fibers cause power degradation of the secondary attacker.
In this letter, three attack scenarios of intrachannel crosstalk accompanied with gain competition attack have been presented and investigated via VPItransMakerTM. We found that the high-powered signal will propagate an intrachannel crosstalk attack to successive three OXCs but with limited two stage switches within each OXC, and this propagation extent depends on switch crosstalk intensity and detection method. The secondary attacker does not have enough attack capability to propagate this attack. We also found if gain competition exists, the system BER will be somewhat higher.
This work is supported by Chang Jiang Scholars Program of the Ministry of Education of China, National Science Fund for Distinguished Young Scholars (No. 60725104), 973 Program (No. 2007CB 310706), National Natural Science Foundation of China (No. 61071101), 863 Program ( 2009AA01Z254, 2009AA01Z215), and NCET Program of MoE of China.